Split out scan call to be explicit to make it easier to refactor duplicate
sections, then move repeated configuration to before/after methods.

Make sure scanning has finished before operations that might concurrently try
to modify the metadata. This also means we can now verify the deletion of the
repository is successful

getJcrSession will open one if needed, which close does not need to do. This
also caused some timing issues in the REST tests as it concurrently operated
on the same session

by placing these in the metadata model / API they can be used more broadly,
while the storage-dependant implementation can be in a plugin. This means the
plugins can be removed from the admin implementation, which can then be used
(as required) in the CLI

use mocks instead of full implementations for testing, cut down dependencies

in case of tests that don't add any repository listeners to the classpath

The test files in artifact converter had been adjusted, and need to merge the
spring context. Also make the artifactConverter injection more specific to
avoid picking up those defined for tests

clean up the dependencies used and add some missing configuration form shade

Should no longer be needed since moving to Maven 3 project API

this avoids modules wanting to report a problem from having to depend on the
plugin itself and its consumer, etc.

avoid duplicating classes in the produced bundle

Large amount of duplicated information and an unnecessary split package
caused duplicates in the produced bundle

Need to line up several dependencies that got managed differently, and trim
out things no longer needed. This could use a diet in general - there should
not be very much of Archiva's API or dependencies needed to build and test a
consumer.

Submitted by Robert David